RobustBench-TC - Misleading Description: leaderboard
Metric: Tool-call accuracy (%) under the MisDesc perturbation (reward component: misleading description on GT; _Budget/_Fast suffix on distractor); RobustBench-TC samples drawn from BFCL V3 single-turn, API-Bank, RoTBench, ToolAlpaca and ToolEyes, each scored by its source benchmark's strict scorer after a format-tolerant tool-call parse, temperature 0 (DeepSeek-R1-Distill at 0.6), Qwen3-family models with thinking disabled; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O4 Mini | 60.8 |
| 2 | DeepSeek R1 Distill Qwen 14B | 42.2 |
| 3 | Qwen 3.5 9B (Non-reasoning) | 41.2 |
| 4 | Qwen 3 32B (Non-reasoning) | 34.3 |
| 5 | Qwen 3 14B (Non-reasoning) | 33.3 |
| 6 | Qwen 3 8B (Non-reasoning) | 32.4 |
| 7 | Qwen 2.5 1.5B Instruct | 22.5 |
| 8 | Qwen 2.5 3B Instruct | 20.6 |
| 9 | Llama 3.2 3B Instruct | 13.7 |
Interactive version: theaggregate.ai/benchmark?slug=robustbench-tc-misleading-description · How It Works · Data refreshed daily, snapshot 2026-10-07.